That's not the hackish way, it's the right way (or at least the only
way)... to address my earlier point, you want to connect icache_port
to an icache and dcache_port straight to the main memory bus.

Steve

On Thu, Jul 22, 2010 at 11:20 AM, Korey Sewell <[email protected]> wrote:
> Someone correct me if Im wrong, but I believe the O3 cache expects 2 ports
> passed to it an "icache_port" and a "dcache_port". So the hack-ish way to do
> it might just be to assign ports to the O3CPU named icache_port and
> dcache_port, but those ports then really connects those ports directly to
> memory instead of a cache.
>
> On Thu, Jul 22, 2010 at 2:02 PM, Steve Reinhardt <[email protected]> wrote:
>>
>> "Easy" is in the eye of the beholder, but it is a change that only
>> involves the python config script.
>>
>> Steve
>>
>> On Thu, Jul 22, 2010 at 9:56 AM, Eberle <[email protected]> wrote:
>> > Is there an easy way to configure the system like that (1 icache only)?
>> > (editing the python script?)
>> >
>> > --
>> > Eberle.
>> >
>> >
>> > On Wed, Jul 21, 2010 at 6:43 PM, Steve Reinhardt <[email protected]>
>> > wrote:
>> >>
>> >> That would probably work.  Thanks for cluing me in... hadn't thought
>> >> about verification applications.
>> >>
>> >> Steve
>> >>
>> >> On Wed, Jul 21, 2010 at 12:48 PM, Eberle <[email protected]> wrote:
>> >> > What about using only an icache and no dcache?
>> >> > We're experimenting with a new verification technique and want to
>> >> > observe
>> >> > the behavior with and without caches.
>> >> >
>> >> > --
>> >> > Eberle.
>> >> >
>> >> >
>> >> > On Wed, Jul 21, 2010 at 2:37 PM, Steve Reinhardt <[email protected]>
>> >> > wrote:
>> >> >>
>> >> >> O3 can deadlock if it doesn't have independent I&D caches... if the
>> >> >> dcache blocks, the pipeline will continuously replay blocked loads,
>> >> >> which saturates the icache port and will prevent the blocked loads
>> >> >> from issuing if they don't have an independent path.
>> >> >>
>> >> >> (At least that's how I remember it... I'm not sure why we go all the
>> >> >> way back to fetch to replay the blocked loads, but it's something
>> >> >> like
>> >> >> that.)
>> >> >>
>> >> >> We never bothered to deal with it because we couldn't think of a
>> >> >> reasonable scenario where anyone would want a detailed CPU model w/o
>> >> >> a
>> >> >> reasonably detailed cache hierarchy too.  I'm curious what your
>> >> >> motivation is...
>> >> >>
>> >> >> Steve
>> >> >>
>> >> >> On Wed, Jul 21, 2010 at 10:06 AM, Eberle <[email protected]>
>> >> >> wrote:
>> >> >> > Is there a way to use the O3 CPU without caches, only using the
>> >> >> > main
>> >> >> > memory?
>> >> >> >
>> >> >> > Thanks,
>> >> >> >
>> >> >> > --
>> >> >> > Eberle.
>> >> >> >
>> >> >> > _______________________________________________
>> >> >> > m5-users mailing list
>> >> >> > [email protected]
>> >> >> > http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
>> >> >> >
>> >> >> _______________________________________________
>> >> >> m5-users mailing list
>> >> >> [email protected]
>> >> >> http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
>> >> >
>> >> >
>> >> > _______________________________________________
>> >> > m5-users mailing list
>> >> > [email protected]
>> >> > http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
>> >> >
>> >> _______________________________________________
>> >> m5-users mailing list
>> >> [email protected]
>> >> http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
>> >
>> >
>> > _______________________________________________
>> > m5-users mailing list
>> > [email protected]
>> > http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
>> >
>> _______________________________________________
>> m5-users mailing list
>> [email protected]
>> http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
>
>
>
> --
> - Korey
>
> _______________________________________________
> m5-users mailing list
> [email protected]
> http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
>
_______________________________________________
m5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users

Reply via email to